In this letter, we propose an autoencoder (AE) for designing Grassmannian constellations in noncoherent (NC) multiple-input multiple-output (MIMO) systems. To guarantee the properties of Grassmannian constellations, the proposed AE constructs the transmitted symbols following an unitary space-time modulation. It penalizes the difference between input and output symbols in terms of cross entropy during the training, which is regarded as a generic optimization method. The constellations learned by the proposed AE have substantial symbol error rate (SER) performance gains compared to the non-Grassmannian constellations and conventionally constructed Grassmannian constellations in high SNR regime. The resulting Grassmannian constellation of the proposed AE achieves higher diversity than the non-Grassmannian constellation in i.i.d. Rayleigh channels. Moreover, the proposed approach can be adaptive to different channel statistics by training with corresponding channel realizations.
翻译:在这封信中,我们提议用一个自动编码器(AE)设计格拉斯曼星座,用于设计非和谐(NC)多投入多产出系统中的格拉斯曼星座;为保证格拉斯曼星座的特性,拟议的AE根据单一时空调制成传送的符号;对培训期间的交叉星座输入和输出符号之间的差别进行处罚,这被视为一种通用优化方法;拟议的AE所学的星座与非格拉斯曼星座和高SNR系统常规建造的格拉斯曼星座相比,具有很大的符号误差率(SER)性能增益;因此,拟议的AE的格拉斯曼星座实现的多样化高于i.d.雷利格河中的非格拉斯曼星座。此外,拟议的方法可以通过对相应的频道实现情况进行培训,适应不同的频道统计数据。